SKA and SKADS Costing The Future

1
SKA and SKADS CostingThe Future

• Paul Alexander
• Andrew Faulkner, Rosie Bolton

2
Costing The SKA

• Our aim is to deliver a final SKA
• No point in design without costing
• Concepts and designs have to fit within a
  realistic budget when costed honestly
• At this stage of the project the priority is
  cost modelling
• Compare and contrast design options
• Identify cost scalings and cost drivers ? modify
  design
• SKADS is playing a major role
• First fully costed design for the SKA Benchmark
  Scenario
• Now coordinating the collaboration between SKADS,
  ISPO team, AU and SA costing teams in a single
  SKA costing exercise and tool whole process
  overseen by ISPO

3
Design and Costing

• SKADS is contributing very significantly to the
  SKA cost model and tool and next generation tools
  and concepts
• SKADS deliverable is a fully costed design
• Excellent progress towards the main SKADS
  deliverable
• Design and costing and evolution of the SKADS
  Benchmark Scenario is a continuing process and
  will continue until at least the end of SKADS
• Rosie Bolton will talk tomorrow about the details
  of the costing process in SKADS
• I will discuss the process and the tools

4
Completed Design and Costing Exercise 1

• Initial DC meeting November 2006
• Work on details of system blocks
• DC Review meeting 15 January 2007
• Preparing written work
• Initial Drafts 9 February
• Consolidating paper
• Final Draft 21 February
• Approval of Paper 28 February
• Publication of Paper 7 March 2007

Revisit for SKA Design Specification Review
5
Needed for stage 2

• Not addresses in SKA Memo 93
• Correlator design and costs
• Central processing hardware, model and costs
• Phased development of Benchmark Scenario costed
• Identify scalings throughout the design
• Better models and methods for
• Dishes
• Accounting practices
• Uncertainty calculations
• Optimisations and tradeoffs

ISPO team collaboration
6
Timetable for Stage 2

• Start second Design and costing exercise 12
  September 2007
• Update those areas not properly analysed in
  stage 1
• ISPO preliminary SKA phase-1 and SKA specs March
  2008
• SKADS input on costed design incorporating AAs
• ISPO final SKA phase-1 and SKA specs March 2009
• Detailed SKADS input

Prepare for SKADS deliverable
7
Costing Methodology

• Ideal is to have a single approach to costing
  within the SKA project
• Share expertise across the entire project
• Have a single tool we can use for comparative
  purposes
• Collaboration between SKADS, ISPO, AU and SA
  costing teams
• Two weeks work meeting in Cambridge in June/July

8
The approaches

• SKADS costing in SKA Memo 93 used a spreadsheet
  approach
• Good exposed costs to everyone involved in the
  project
• Good easy to use and modify by non-experts
• Good mapped costing onto the design elements of
  the telescope and kept the model intimately
  linked to the design
• Bad difficult to simulate, do montecarlo
  uncertainty modelling
• Bad consistency and error checking done by hand
• Bad lacks flexibility and poor scaling of
  elements of the design

9
The approaches

• ISPO modelling, SKA Memo 92
• Good simulation environment with montecarlo
  modelling
• Good scaling and model exploration
• Good consistency and error checking
• Bad difficult for non experts to alter
• Bad assumptions not easily exposed to the user
  the tool can be dangerous if the underlying model
  is wrong or not appreciated by the users
• Bad difficult to incorporate a real design or
  consider effects of design changes
• SA costing for MeerKAT
• Good used same original code base as ISPO
• Good excellent tools to visualise outputs
• Bad see ISPO

10
Structure of a new tool
SKA realisation
Design Block
Design Block
Design Block
Design Block
Design Block
Component
Component
Component
Component
11
Structure of the new tool

• Adopt the SKADS high-level view
• System consisting of Design Blocks we combine
  together to form a realisation of the telescope
• Design blocks use costed components and/or cost
  models which are determined by domain experts
• The costing process involves everyone
• Use a modified ISPO code base for the
  implementation new UI
• Python code (initially) implements the
  specifications / design of each design block
• Standard software tools provided to ensure
  consistency, error checking, uncertainty
  modelling
• GUI enables non-experts to browse, modify and
  develop components and design blocks (some python
  code needed)
• Build in the MeerKAT approach to visualisation

12
Design Blockreporting and design structure
Cost modelling code(python or )
Library of DBs and components(xml format)
GUI interface
13
Structure of the new tool
14
Example Model Aperture Array
15
GUI Snapshot
16
The Repository

• Develop libraries of components and design blocks
  in a repository
• Domain experts review, extend and modify these
• Expect an SKA wide repository and also
  repositories local to other projects one at least
• Individuals can have their own library or modify
  existing system realisations, design blocks and
  components

17
Using the Model

• Astronomer
• Use an existing design of the SKA and investigate
  cost and performance tradeoffs within the design,
  output performance information/statistics
• System designer or expert radio astronomer
• Construct SKA realisations (systems) out of
  existing design blocks and components
• Engineer
• Develop and review design blocks and components

18
Timeline Best Efforts

• End September
• Include basic AA description into costing tool
• Mid October
• GUI to browse component library
• Underlying change to system architecture
• Mid November
• GUI to browse / modify design block library
• January/February 2008
• First release of an integrated tool

19
Conclusions

• Initial design and costing has been a very good
  first step towards our final deliverable
• We have learnt a great deal about feasible and
  affordable designs
• SKADS coordinating intercontinental effort to
  deliver the SKA costing tool
• Work to date crucial in establishing AA
  technology as a core technology for all SKA
  options